]
Nicky Mølholm commented on WFLY-3761:
-------------------------------------
Hi [~dlofthouse] - I recently created a security related community discussion regarding
JMS's authentication in Wildfly (
)
It turned out i needed to send a certain plaintext property from the client before it
began to work successfully. Again here the behavior was utterly complete silence.
Also, in my company's real-world codebase we faced the same complete silence because
the "Custom Login module in our security domain" had classloading issues when
used from the remoting subsystem (through a jaas based security realm).
I mention these "war stories" too - to supply with my personal perspective that:
* the silent security behavior in Wildly is making it really difficult for us to get
"customized security mechanisms" right.
It is by no means a show stopper. But it is really really troublesome from time to time
when one doesn't get the entire end-to-end security configuration correct the first
time.
Security realms does not validate JAAS references to security
domains
---------------------------------------------------------------------
Key: WFLY-3761
URL:
https://issues.jboss.org/browse/WFLY-3761
Project: WildFly
Issue Type: Bug
Security Level: Public(Everyone can see)
Components: Domain Management, Security
Affects Versions: 8.1.0.Final
Environment: Development Mac
Test Linux (Debian)
Reporter: Nicky Mølholm
Labels: jaas, logging, security, trace
Fix For: Awaiting Volunteers
*Problem*
In the server configuration file (standalone.xml) it is possible to define a security
realm that points to a security domain that does not exist - and there is no error
reporting of this at all. There is no trace information of this at all, either.
*Example*
* Download a stock Wildfly 8.1.0.Final
* Replace standalone.xml with this gist:
https://gist.githubusercontent.com/nickymoelholm/4908092afdcd519361df/raw...
Run it and you will see now errors at all. Despite the fact that the _FlawedRealm_ points
to a bogus security domain called _ThisDomainDoesntExistAtAll_ . I have captured my
logoutput too. Find it here:
https://gist.githubusercontent.com/nickymoelholm/4908092afdcd519361df/raw...
*What is wrong with this behavior?*
The bootstrapping process must validate that the configuration is valid indeed. It really
doesn't - not semantically that is. Only XSD compliance / XML syntax wise. And if, for
some weird reason, that silence is "security" - then at least let us know of the
errors on loglevel = TRACE.
*Why is this issue created?*
The silent behavior makes security configuration in Wildfly an _extremely expensive
operation_ in terms of time spent by the average Java EE developer / administrator. I have
created this issue because I want wildfly to help developers/administrators become better
at spotting our errors - because, in the end, that is a tangible productivity booster.